Game Theory


    3.3 Game Theory
    Game theory is a branch of applied mathematics, which deals with multi-person decision making situations. The basic assumption is that the decision makers pursue some well-defined objectives and take into account their knowledge or expectations of the other decision makers’ behavior. Many applications of game theory are related to economics, but it has been applied to numerous fields ranging from law enforcement [19] to voting decisions in European Union [20].
    There are two main ways to capitalize game theory. It can be used to analyze existing systems or it can be used as a tool when designing new systems. Existing systems can be modeled as games. The models can be used to study the properties of the systems. For example, it is possible to analyze the effect of different kind of users on the system. The other approach is implementation theory, which is used when designing a new system. Instead of fixing a game and analyzing its outcome, the desired outcome is fixed and a game ending in that outcome is looked for. When a suitable game is discovered, a system fulfilling the properties of the game can be implemented.
    Most game theoretical ideas can be presented without mathematics; hence we give only some formal definitions. Also, introduce one classical game, the prisoner’s dilemma which we use to demonstrate the concepts of game theory.
    In the prisoner’s dilemma, two criminals are arrested and charged with a crime. The police do not have enough evidence to convict the suspects, unless at least one confesses. The criminals are in separate cells, thus they are not able to communicate during the process. If neither confesses, they will be convicted of a minor crime and sentenced for one month. The police offer both the criminals a deal. If one confesses and the other does not, the confessing one will be released and the other will be sentenced for 9 months. If both confess, both will be sentenced for six months. The possible actions and corresponding sentences of the criminals are given in Table 3.1.
    Criminal 2
    Don’t confess
    Confess
    Criminal 1
    Don’t confess
    -1,-1
    -9,0
    Confess
    0,-9
    -6,-6
                                    Table 3.1: Prisoner’s dilemma
    3.3.2 Assumptions and Definitions
                    Game: A game consists of players, the possible actions of the players, and consequences of the actions. The players are decision makers, who choose how they act. The actions of the players result in a consequence or outcome. The players try to ensure the best possible consequence according to their preferences.
                    The preferences of a player can be expressed either with a utility function, which maps every consequence to a real number, or with preference relations, which define the ranking of the consequences. With mild assumptions, a utility function can be constructed if the preference relations of a player are known [21].
                    Rationality: The most fundamental assumption in game theory is rationality. Rational players are assumed to maximize their payoff. If the game is not deterministic, the players maximize their expected payoff. The idea of maximizing the expected payoff was justified by the seminal work of von Neumann and Morgenstern in 1944 [21].
                    The rationality assumption has been criticized. Experiments have shown that humans do not always act rationally [22]. In telecommunications, the players usually are devices programmed to operate in a certain way, thus the assumption of rational behavior is more justified.
                    The maximizing of one’s payoff is often referred to as selfishness. This is true in the sense that all the players try to gain the highest possible utility. However, a high utility does not necessarily mean that the player acts selfishly. Any kind of behavior can be modeled with a suitable utility function. For example, a preference model called ERC [23] not only pays attention to the benefit of the player, but also the benefit relative to the other players. In many occasions, an ERC model fits experimental data better than simpler models, where the players only try to maximize their own benefit.
                    It is also assumed that the players are intelligent, which means that they know everything that we know about the game and they can make the same deductions about the situation that we can make.
                    Solution: In game theory, a solution of a game is a set of the possible outcomes. A game describes what actions the players can take and what the consequences of the actions are. The solution of a game is a description of outcomes that may emerge in the game if the players act rationally and intelligently. Generally, a solution is an outcome from which no player wants to deviate unilaterally.
                    Pareto Efficiency: An outcome of a game is Pareto efficient, if there is no other outcome that would make all players better off. In the prisoner’s dilemma, all the outcomes except (Confess; Confess) are Pareto efficient. In the battle of the sexes, the outcomes in which both attend the same event are Pareto efficient. In implementation theory, the aim is typically to design a game that will end in a Pareto efficient outcome.
                    Pure and Mixed Strategies: When a player makes a decision, he can use either a pure or a mixed strategy. If the actions of the player are deterministic, he is said to use a pure strategy. If probability distributions are defined to describe the actions of the player, a mixed strategy is used. For example, in the battle of the sexes the husband can choose the hockey match with a probability of 70 percent. If mixed strategies are used, the players maximize their expected payoff.
    3.3.2.1 Classification of Games
                    Games can be classified into different categories according to their properties. The terminology used in game theory is inconsistent, thus different terms can be used for the same concept in different sources.
                    Defective and cooperative games: Games can be divided into defective and cooperative games according to their focus. Cooperative games are also called coalition games. In defective games, the actions of the single players are considered. Correspondingly, in coalition games the joint actions of groups are analyzed, i.e. what is the outcome if a group of players cooperate? The interest is in what kind of coalitions form. The prisoner’s dilemma is defective games.
                    In telecommunications, most game theoretic research has been conducted using defective games, but there are also approaches using coalition games. Coalition games can be used to analyze heterogeneous ad hoc networks. If the network consists of nodes with various levels of selfishness, it may be beneficial to exclude too selfish nodes from the network if the remaining nodes get better quality of service that way.
                    Strategic and extensive games: In strategic or static games, the players make their decisions simultaneously at the beginning of the game. While the game may last long and there can be probabilistic events, the players can not react to the events during the game. The prisoner’s dilemma and the battle of the sexes are both strategic games.
                    On the other hand, the model of an extensive game defines the possible orders of the events. The players can make decisions during the game and they can react to other players’ decisions. Extensive games can be finite or infinite. Formal definitions of strategic and extensive games are given later.
                    A class of extensive games is repeated games, in which a game is played numerous times and the players can observe the outcome of the previous game before attending the next repetition. A typical example is a repeated prisoner’s dilemma in which the same situation is repeated several times.
                    Zero-sum games: Games can be divided according to their payoff structures. A game is called zerosum game, if the sum of the utilities is constant in every outcome. Whatever is gained by one player is lost by the other players. Gambling is a typical zero-sum game. Neither of the example games are zero-sum games. Zero-sum games are also called strictly competitive games. In telecommunications, the games are usually not zero-sum games. However, if a simple scenario, for example the bandwidth of a single link, is studied, the game may be a zero-sum game.
                    Games with perfect and imperfect information: If the players are fully informed about each other’s moves, the game has perfect information. Games with simultaneous moves have always imperfect information, thus only extensive games can have perfect information.
                    A game with imperfect information is a good framework in telecommunications, because the users of a network seldom know the exact actions of the other users. However, it is often more convenient to assume perfect information.
                    Games with complete and incomplete information: In games with complete information the preferences of the players are common knowledge, i.e. all the players know all the utility functions. In a game of incomplete information, in contrast, at least one player is uncertain about another player’s preferences.

                    A sealed-bid auction is a typical game with incomplete information. A player knows his own valuation of the merchandise but does not know the valuations of the other bidders.

                                                                                                                                      Order Now